How to aggregate Attribute Weight results and sum them up?
I want to sum up all attribute weights for my attributes (e.g in a data table / example set), to see which ones are really important by "majority vote"
how can I do that? cant do it with aggregate or append, it complains that it not a data table, and there seems no way to get the columns "weight" or attribute weight directly, e.g from a log..
Best Answers
-
Telcontar120 RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,635 Unicorn
On the menu bar if you click on "file" and then "new process" you'll have the option to choose one of the existing templates. You want the "predictive maintenance" template. See the attached screenshot.
In that example process there is a subprocess that has been labeled "determine influence factors" which has a bunch of processing related to using weights that you will prboably find helpful.
0 -
Fred12 Member Posts: 344 Unicorn
hi,
ok I looked into that process, and its quite understandable the "determine influence factors" subprocess...
somehow similar to what I did... however I was not impressed by the ranking of the attributes,
I excluded the first 2 most important attributes with select attributes operator, and it didnt change anything in the performance in the log... only when I excluded the first 5 most important sensors, performance changed by 0.027....
0
Answers
hi Fred,
there is a weight to data operator that will convert the weight table to exampleset
then you can use hte aggregate operator to sum it up.
If instead of sum you are looking for other way to summarize check the predictive maintenance template that comes with RapidMiner, Look for the "Determine influence factors" there is provides an example of averaging but more fancier
I couldnt find any operator "Determine influence factors" .
do I have to connect each "Weights to Data" and then append them together? ...